end-of-text token between documents
A neuronOne of thousands of tiny dials inside each layer. It reads the text so far and produces a number, its activation, that nudges what the model writes next. in layerOne step in the stack the text passes through. Qwen3-0.6B has 28 layers, each with 3,072 neurons. 27 of Qwen3-0.6B. Its strongest snippets are very consistent: they mostly peak on the same token. It is featuredA neuron picked for the full treatment: steering outputs and a written explanation. About 3,000 of the 86,000 neurons are featured., so it has steering outputs.
What this neuron does
This neuron fires on the hidden marker that separates one document from the next, much like its neighbours. Turning it up made the model repeat single words such as 'football', 'music' and 'cloudy', which is not clearly related to the trigger.
Written by Claude from this neuron's activations and steering outputs. Confidence: medium. Think it is wrong? Flag it (opens a prefilled issue on GitHub).
Fires most on
The text windows where this neuron's activationThe number a neuron produces for a token. Large positive means it fired hard; negative means it pushed the other way. was highest. Amber marks positive activation; the outlined tokenA piece of text the model reads at once: often a word, sometimes part of a word or a punctuation mark. is the peak. Hover a token for its strength.
-
/definition/acanthosis nigricans>.<|endoftext|>Dronacharya was the teacher of both the Kauravas and Pand
-
ycetica<|endoftext|>Adding Fractions Level 3⏎The 5-step planIn Adding Fractions
-
-Pecos.<|endoftext|>top of page⏎Patterns of tree-liana interactions: distribution and host preference of
-
to like.<|endoftext|>The late Dominica-born writer Jean Rhys, best known for her novel “
-
edible nut.<|endoftext|>Canal Park is a little tourist area around the canal that leads into the Saint
-
criteria.<|endoftext|>The haunting image of the fallen soldier, aged just 18, is one
SteeringAdding a push along a neuron while the model writes, to see how its output changes. Turning it up or down.
Push this neuron up or down while the model writes, and watch what changes. The outputs were generated ahead of time, so nothing runs in your browser.
Here is something I have been thinking about lately: football football football football football football football football football football football football football football football football football football football football football football football football football football football football football football football football football football football football football football football football football football football football football football football football
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
realize tradition is a living and operative reality. Thinking about the experiences of the French in the midst of the Vichy regime and the Second World War, Simone Weil wrote:⏎“In the atmosphere of anguish, confusion, solitude, uprooted
-
agonists (Allbuterol/Salbutamol, Proventil, Ventolin) and theophylline (Theodur, QTSR) work by opening up the bronchial tubes
-
here ensures a smooth and easy understanding of all the concepts. Understand the concepts behind every Unit and score well in the board exams.⏎6. How does the play The Duchess of Malfi end? Do you suggest any alternative ending?⏎Ans:
-
right of free speech does not exist.⏎Iocasta: That's a slave's life—to be forbidden to speak one's mind.⏎(Euripides, The Phoenician Women; author's translation)⏎Writers in exile often face the
Connections
Show on mapWhich neuronsOne of thousands of tiny dials inside each layer. It reads the text so far and produces a number, its activation, that nudges what the model writes next. in earlier layersOne step in the stack the text passes through. Qwen3-0.6B has 28 layers, each with 3,072 neurons. feed this one, and which later ones it feeds. Amber links excite, blue links suppress; longer bars are stronger.
Fed by earlier layers
- L26:1564 strongly excites
- L25:11 suppresses
- L26:2 excites
- L22:201 excites
- L8:14 excites
- L8:16 suppresses
- L6:8 suppresses
- L24:86 suppresses
- L21:403 suppresses
- L10:119 weakly excites
- L8:7 weakly excites
- L6:13 weakly excites
- L8:4 weakly excites
- L18:1288 weakly excites
- L11:49 weakly excites
- L0:1301 weakly suppresses
- L0:1224 weakly suppresses
- L26:732 weakly excites
- L17:139 weakly suppresses
- L10:106 weakly excites
- L18:214 weakly excites end of document marker
- L0:178 weakly suppresses
- L21:1227 weakly suppresses
- L21:137 weakly suppresses
Feeds later layers
None: this is the last layer.
Activation histogramA chart of how often a neuron takes each activation value.
How often this neuron fires at each strength. Almost always it does close to nothing (the tall bars in the middle); the snippets above are the rare moments it fired hard, amber for positive and blue for negative.
Logit lensReading which output tokens a neuron pushes toward or away from, straight off its output weights, before the later layers have their say.
The words this neuron nudges the model toward (amber) or away from (blue) when it fires. A hint, not a label: many neurons work through later layersOne step in the stack the text passes through. Qwen3-0.6B has 28 layers, each with 3,072 neurons. instead.
Promotes
- 一是
- (!_
- �
- (({
- []{"
- {{$
- việc
- �
- <?=
- (!_
Suppresses
- 冰淇淋
- Bicycle
- ambulance
- Football
- 赌博
- SHIP
- Soup
- football
- Miner
- 饮料